Coevolution of farming and private property during the

early Holocene
Samuel Bowles
a,1
and Jung-Kyoo Choi
b
a
Santa Fe Institute, Santa Fe, NM 87501; and
b
Kyungpook National University, School of Economics and Trade, Daegu 702-701, Korea
Edited by Bruce P. Winterhalder, University of California, Davis, CA, and accepted by the Editorial Board April 12, 2013 (received for review July 16, 2012)
The advent of farming around 12 millennia ago was a cultural as
well as technological revolution, requiringa newsystemof property
rights. Among mobile huntergatherers during the late Pleistocene,
food was almost certainly widely shared as it was acquired. If a har-
vestedcrop or the meat of a domesticatedanimal were tohave been
distributed to other group members, a late Pleistocene would-be
farmer would have had little incentive to engage in the required
investments in clearing, cultivation, animal tending, and storage.
However, the new property rights that farming requiredsecure
individual claims to the products of ones laborwere infeasible
because most of the mobile and dispersed resources of a forager
economy could not cost-effectively be delimited and defended. The
resulting chicken-and-egg puzzle might be resolved if farming had
been much more productive than foraging, but initially it was not.
Our model and simulations explain how, despite being an unlikely
event, farming and a new system of farming-friendly property
rights nonetheless jointly emerged when they did. This Holocene
revolution was not sparked by a superior technology. It occurred
because possession of the wealth of farmerscrops, dwellings,
and animalscould be unambiguously demarcated and defended.
This facilitated the spread of new property rights that were advan-
tageous to the groups adopting them. Our results thus challenge
unicausal models of historical dynamics driven by advances in tech-
nology, population pressure, or other exogenous changes. Our ap-
proach may be applied to other technological and institutional
revolutions such as the 18th- and 19th-century industrial revolution
and the information revolution today.
agent-based simulation
|
evolutionary game theory
|
technical change
|
institutional change
|
big history
A
mong the outstanding puzzles in human social dynamics is
the simultaneous emergence and diffusion of novel economic
institutions and new technologies. An important example is the
appearance of private storage (14) and other indicators of in-
dividual property (5, 6) along with the advent of farming (7, 8),
one of the greatest technologicalinstitutional revolutions ever
experienced by our species.
Economists, archaeologists, and historians often consider
advances in technology to be an exogenous driver of innovations in
property rights, wealth inheritance practices, or other institutions.
Well-studied examples include the social and cultural impacts of
the introduction of the horse on the US Great Plains in the 17th
century or the sweet potato in the New Guinea Highlands a cen-
tury later (9, 10). However, the technological and institutional
innovations of the early Holocene are difcult to reconcile with
this view because, as a number of archaeologists have pointed out
(1113), farming was probably not economically advantageous in
many places where it was rst introduced. Indeed, recent estimates
suggest that the productivity of the rst farmers (calories per hour
of labor including processing and storage) was probably less than
that of the foragers they eventually replaced, perhaps by a con-
siderable amount (14) (SI Appendix). In many parts of the world,
stature and health status appear to have declined with cultivation
(15). Farming did raise the productivity of land and animals, and
this, we will see, was critical to its success. However, why an erst-
while huntergatherer would adopt a new technology that in-
creased the labor necessary to obtain a livelihood remains a puzzle.
The other often-proposed exogenous driver of the Holocene
revolutionpopulation pressure (16)fares no better. What is
thought to be the rst independent emergence of farmingin the
Levantfollowed almost 8 centuries of population decline (17).
The archaeological record is thus inconsistent with the idea that
under the more favorable Holocene weather conditions, farming,
once invented, was simply a better way to make a living, and that
the new property rights subsequently emerged in response to the
needs of the new technology.
How, then, did this new technology and novel system of prop-
erty rights emerge and proliferate? We propose that the new
property rights and the new way of making a living coevolved,
neither being viable alone but each providing the conditions
permitting the advance of the other. This coevolution hypothesis
is based on two empirically motivated premises: that farming re-
quired a novel system of property rights, and that (in the absence
of exceptional circumstances) this system of farming-friendly
property rights was not viable in an economy based on wild plant
and animal species. This is why coevolution was possible and in-
dependent evolution unlikely.
Fig. 1 contrasts the causal structure of our coevolutionary model
with the technology-driven approach. The vertical causal arrows
(Fig. 1, Right) make it clear that farming was essential to the advent
of novel institutions; however, in our interpretation, neither its
advent nor its contribution to the spread of new property rights
derived from any initial advantage in the productivity of labor.
In the next section, we explain the causal structure of our model,
showing how farming and the extension of private property rights
each provided conditions favorable for the proliferation of the
other. We then represent this coevolutionary causal structure in a
mathematical model. This allows an analysis of the congurations
of technologies and institutions that could persist over long periods,
as well as the process of transitionfroma congurationrepresenting
a huntergatherer economy to one representing farming with its
associated property rights.
The test of the causal explanation expressed by the model is
whether, when calibrated to represent late Pleistocene and early
Holocene conditions, our simulations of the model reproduce the
basic known facts about the timing, nature, and diversity of the
transition process. These facts include not only the Holocene
transition (where it occurred) but also the very long term persis-
tence of foraging not only before the Holocene, but in many parts
of the world extending right up to the middle of the last millen-
nium. This is the test that we set for ourselves in the subsequent
two sections, rst using climate, archaeological, and other data to
calibrate the model, and second to check whether the simulations
do indeed replicate what is known about the emergence of
farming and the novel institutions associated with it.
Author contributions: S.B. and J.-K.C. designed research, performed research, analyzed
data, and wrote the paper.
The authors declare no conict of interest.
This article is a PNAS Direct Submission. B.P.W. is a guest editor invited by the
Editorial Board.
Freely available online through the PNAS open access option.
1
To whom correspondence should be addressed. E-mail: samuel.bowles@gmail.com.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
1073/pnas.1212149110/-/DCSupplemental.
www.pnas.org/cgi/doi/10.1073/pnas.1212149110 PNAS Early Edition | 1 of 6
E
C
O
N
O
M
I
C
S
C
I
E
N
C
E
S
In addition to providing an alternative to the exogenous tech-
nology- or population-driven interpretation, our method is also
unconventional in the study of prehistory. Evolutionary game the-
ory combined with agent-based simulation allows us to track how,
as a result of independent individual actions, a population may
come to engage in new livelihoods and institutions. As a result, we
can study a population when it is in motion, not only when it is
stationary, allowing us to capture processes rather than focus on
end points, or before vs. after scenarios. This contrasts with
models in which entire groups are represented as though they were
a single individual, so that group differences in livelihoods or
institutions are an all-or-nothing matter. In our simulations, not
everyone conforms to the same institutional norms and groups
virtually always pursue mixed livelihoods, some mostly foraging,
others with a majority of farming.
Mutual Dependence of Farming and Farming-Friendly
Property Rights
Systems of property rights govern access to the things that people
value, and they differ in who can legitimately exclude whom from
what. These differences are often both subtle and complex, so we
will represent the property rights of a group not categorically but
as a continuum based on the diverse behaviors of the individuals
making up the population. We simplify the multidimensional
nature of property by letting each individual adopt either com-
mon property or private property based on individual possession.
The latter (henceforth private property) gives the holder (an
individual or nuclear family, hence the term private) the right
to determine the use of the resource that is owned, including, of
course, the right legitimately to exclude others from its use.
Rights in this sense convey an expectation that their exercise will
not generally be legitimately or successfully opposed by others.
Common property pertains to goods that are not covered by
private property and instead cannot be legitimately monopolized by
a single individual (or family). Common property does not entail
equality in the use of resources, but it does prohibit monopolization
of a resource by an individual or family. Under common property,
the rights of use or exclusion are exercised by groups larger than
a single family, as when the members of a foraging band exclude
nonmembers from its territory or when a lineage claims territori-
ality over a shing site.
If widely adopted in a group, a particular set of property rights
may benet its members when it accomplishes two things: pro-
viding incentives for the productive use of labor and other scarce
inputs, and avoiding costly conicts among group members. Which
set of property rights best accomplishes these endsprivate,
common, or some admixturedepends on the nature of the goods
and services making up the livelihood of a group. The property
rights that worked well in a foraging economy were quite different
from those adapted to farming.
Our rst premisefarming required private propertyis illus-
trated by the barriers to cultivation in the absence of appropriate
property rights encountered by two would-be farmers, members of
a group of foragers in Malaysia, the Batek: The traditional Batek
notions that all natural resources are unowned until collected and
that any food obtained in excess of the needs of the procurers
family must be shared with other families seem well suited to
a nomadic foraging life, but wholly unsuited to ... farming (18).
The two Batek men who had discovered cultivated rice tried
planting some. However, their fellow group members simply har-
vested it (and, of course, felt obliged to share the harvest with the
entire group). Similar cases of free riders claims on would-be rst
farmers are found among the !Kung in southern Africa and the
Hiwi in Venezuela (19). James Woodburn concluded that the
value systems of non-competitive, egalitarian hunter-gatherers
limit the development of agriculture because rules of sharing re-
strict the investment and savings necessary for agriculture (20).
The delayed returns from long-term investments in food pro-
duction (both cultivation and animal tending) would not have been
enjoyed by the rst farmers in the absence of property rights en-
suring that a family could effectively claim the food and other
subsistence goods that they produced, excluding others. However,
although foragers typically own some ornaments, tools, and other
valued items, and entire groups of foragers sometimes delimit and
defend hunting and gathering territories, the absence of effective
storage combined with the need to smooth out the day-to-day va-
garies of individual food procurement dictates that food be widely
shared as it is acquired. (By foragers, we mean mobile hunter
gatherers; we will see that sedentary huntergatherers are an ex-
ception that may account for one of the best-documented cases of
the Holocene coevolution of farming and private property.)
Consumption smoothing through food sharing is practiced in
all small-scale societies, and practices among forager groups differ
(21). However, available data on recent and contemporary for-
agers and horticulturalists show that the fraction of the food one
has acquired that is subsequently transferred to other households
is substantial among foragers and is much greater than among
hand-technology farmers (SI Appendix). An individual hunter may
have the right to distribute his prey to others, but monopolizing it
for his immediate family would typically be a serious transgression
of social norms and would undermine the system of mutual in-
surance based on food sharing.
Widespread sharing of food among foragers would have been
particularly pronounced during the extraordinarily volatile cli-
matic conditions of the late Pleistocene. Thus, it seems likely that
when climate variability ameliorated at the end of the Pleistocene,
farming-friendly private property rights over the means of sub-
sistence produced by ones labor were for the most part absent
among foragers.
Our second premiseprivate property required farming
suggests a reason for this absence: Demarcating and enforcing
individually held property rights were not cost-effective (or
sometime even possible) in the diffuse and in many cases mobile
wild species on which foragers subsisted. The exceptionsgroup
rights in especially concentrated resource patches such as migra-
tory routes of prey or the dwellings of sedentary foragersare
suggestive of how farming provided a more favorable environment
for the property rights it required. What made the defense of
concentrated resource patches feasible and cost-effective was that
they were highly productive (in value per unit of space), so that
a limited investment in demarcation and the exclusion of others
would be justied by the substantial losses that these expenditures
would prevent (22). Similarly, the concentrated and easily looted
resources of a farming economy made possible by the associated
increase in the productivity of land and animalscrops ready for
harvest or in stores, as well as livestockwere worth defending.
Thus, farming provided favorable conditions for the new property
rights that it required.
Where the Holocene revolution occurred, acknowledgment of
ones fellow rst farmers property rights in dwellings, crops, and
animals became a convention, that is, a norm that persisted once
it had become common because respecting the rights of others
was in the interest of each as long as two conditions were met.
The rst was that most people accept the exclusion of others
from ones possessions as legitimate. The second was that the
things one valued could be made to be possessions, that is, they
Resource rich hunting &
gathering environments
c
Individual possession-
based property in the
means of subsistence
d
Improved climate
a
Farming
b
Sedentism
Fig. 1. Holocene coevolution of farming and private property. A conven-
tional causal sequence is that a is a sufcient condition for b, which is then
followed by d. Our model and simulations suggest a coevolutionary scenario:
In the absence of farming, c is a necessary condition for d, which in con-
junction with a provided the necessary but not sufcient conditions for b.
2 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1212149110 Bowles and Choi
could be unambiguously demarcated and then defended so that
contests over the facts of possession would be infrequent.
Whereas the facts of possession were unlikely to be contested
when wealth took the form of domesticated animals and cereals
stored within ones dwelling, in a foraging economy the second
condition was unlikely to be met. As a result, individuals asserting
private property rights in a foraging economy would not only be
challenged by those upholding an alternative common property
rule, they would even challenge one another over the facts of
possession. The property rights that farming thus allowed did
more than provide individual incentives for the investments that
farming required. Where farming was introduced and where most
people adhered to the norm of respecting others possessions,
private property attenuated costly within-group conicts, allowing
populations of the rst farmers to prosper even if farming was not
initially more productive than foraging.
Combining the two premises, it is easy to see that the farming
private property match, once established, could persist, just as the
forager-common property match had persisted for tens of mil-
lennia. But how could the new institutions emerge in the rst
place? Unlike a newcrop or method of cultivation, a systemof new
property rights could not be introduced piecemeal by a single in-
dividual. New institutionsnovel property rights includeddo
not work well unless most members of a community adhere to
them. Successful adoption of farming-friendly property rights thus
required a critical mass. This critical mass problem could, in
principle, be overcome by a governmental at introducing and
enforcing a new system of property for an entire population.
However, property rights appropriate for farming emerged and
proliferated many millennia before there were governments that
could enforce them, and any single individual initially asserting
such rights would most likely have been unsuccessful, as the hap-
less Batek would-be rice farmers discovered.
In light of this critical mass problem and the likely absence of
a productivity advantage for the rst farmers, it is difcult to explain
howeither farming or the newproperty rights that farming required
could have rst emerged and then proliferated at the end of the
Pleistocene, when both were rare. Emergence and subsequent
proliferation are two distinct puzzles. The rst requires an expla-
nation of how a single group surviving entirely on wild resources
could have taken up food production, and the second, an expla-
nation of why, even in the absence of a labor productivity advan-
tage, the farming practices of such a group would have spread
throughout an ethnolinguistic unit, rather than being reversed.
It is not difcult to explain cases where, following its emergence
and full development in one location, the farming-cum-private
property package was introduced to a new areaas in the case of
the importation of farming and herding to Cyprus around 10,500
B.P., or its expansion to many parts of Europe (23, 24). For this
reason, we will focus on explaining the handful of well-studied
cases of the independent emergence and local adoption of
farming and its associated property rights.
Modeling the Coevolution of Technology and Institutions
No single account can capture the distinctive trajectories by
which this occurred during the early Holocene (7, 8, 2429). The
conditions under which farming and farming-friendly property
rights jointly emerged differed in many respects: the crops and
animals that were cultivated and eventually domesticated, the
environmental conditions under which this occurred, the degree
of long-term investment required, and the nature of property
rights best adapted for their use. We think, nonetheless, that
there may be some common causal mechanisms underlying the
Holocene revolution. An adequate model should provide the
causal mechanisms accounting for what is known from archaeo-
logical, biological, and other evidence.
Enduring transitions occurred no more than 12,000 y ago
and they were very rare; in most ethnolinguistic units, the farming-
cum-private property package did not independently emerge.
Transitions were slow and sometimes witnessed reversals. The
passage from initial domestication of one or two species ac-
counting for a modest portion of the diet to a primary commitment
to food production in some cases extended over as many as six
millennia. As a result, mixed societies with substantial portions of
the diet coming from both farming and huntinggathering per-
sisted over long periods (30).
We provide a model of cultural evolution that captures the
complexity and diversity of this process and allows us to identify the
underlying causal mechanisms accounting for the joint emergence
of the newtechnology and property rights. The model, described in
SI Appendix, illuminates the population dynamics resulting from
the causal mechanisms we have described. We modeled individu-
als choices of both technology and behavior toward other group
members. We then used the uctuations in surface temperature
over the past 40,000 y as an indicator of climatic volatility, along
with estimates of migration, intergroup competition, and other
aspects of the demography of late Pleistocene and early Holocene
small-scale society to calibrate our model so as to capture the
conditions under which the new individual strategies of ownership
and procuring a livelihood might have proliferated.
Our model represents the social and technological dynamics of
a population similar in size and composition to a late Pleistocene
ethnolinguistic unit (600 individuals per generation), made up of
many partially isolated subpopulations (called groups, of 20
individuals per generation) about the size of forager bands or
small villages. In this model, farming and private property spread
as a result of adoption by most individuals in a group occurring
either as the result of changes within the group or from emula-
tion by a group of foragers and their subsequent adoption of the
new institutions and technology.
There are three stages in the model: production, distribution, and
cultural updating. In the production stage, individuals adopt one of
two technologies: farming or huntinggathering. These technolo-
gies differ in four ways: Foraging is more productive than farming;
farming requires a prior investment that may be lost if the product
is contested; the farming product is more readily demarcated and
defended than the foraging product; and because of its sedentary
nature, farming is more disadvantaged by volatile weather.
In the distribution stage that follows production, independent of
the choice of technology, each individual interacts with a randomly
paired other member of their group to divide their products
according to a modied version of Maynard Smiths bourgeois
hawkdove game (31). The strategies that an individual may adopt
reect patterns of sharing, aggrandizement, and collective disci-
pline found in ethnographic studies of huntergatherers and hor-
ticulturalists (3236). Similar to the dove in the hawkdove game,
the rst behavioral type, the sharer, concedes half of the product to
the other, or the whole product if the other claims it. Bourgeois
individuals (the second behavioral type) claim the entire product if
it is in their possession. A nonpossessing bourgeois may engage in
contests with another bourgeois if possession of the product is
contestable, which is less likely for farmed products. A key aspect
of our model is that bourgeois behavior will differ markedly
depending on the technology in use. Because farming wealth is
readily demarcated and defended, bourgeois individuals respect
the farmed possessions of others, whereas for foraged products they
sometimes claim the possessions of others, resulting in conicts.
The third behavioral type, the civics, act exactly like sharers
when they meet a fellow civic or a sharer (they share), but when
paired with an individual who refuses to share (e.g., a bourgeois
in possession of the product), they join with any other civics in the
group to contest the claims of the bourgeois, succeeding with
a probability that is increasing with their numbers (SI Appendix,
Fig. S2). If they succeed, they distribute the bourgeois product
among all the civics, whereas the losing bourgeois bears a cost. If,
instead, the civics fail, they bear the losers cost. [The civic strategy
is based on ethnographic studies of the maintenance social order
in stateless societies (36, 37) so as to allow a more realistic co-
ordinated rather than individual punishment process (38).]
Bowles and Choi PNAS Early Edition | 3 of 6
E
C
O
N
O
M
I
C
S
C
I
E
N
C
E
S
In addition to within-group individual interactions, groups of
the same ethnolinguistic unit interact in what may be termed
emulation contests. These could be outright violent conicts or
simply the encroachment by groups with higher payoffs on less
successful groups. The key is that the group with the higher total
payoff becomes a cultural model for the less successful group.
Losing groups cede some resources to the winning group and are
then culturally assimilated by the winners (explained below). In
each generation, there is a probability that each individual will
migrate randomly among groups.
In the cultural updating stage, individuals are paired with
a cultural model (for example, an elder) who is more likely to be
from the numerically predominant type in the group than would
occur by chance (SI Appendix, Fig. S3), reecting conformist
cultural transmission (39). If the two are the same type, no
updating occurs. However, if they differ, and if the models payoff
in the previous period was higher than the updating individuals,
he or she switches to the type of the model. In groups that have not
lost a contest with another group, the cultural model is selected
fromthe groups own population, whereas in groups that have lost
a contest in the previous period, updating individuals are paired
with models drawn from the winning group, thereby tending to
spread those technologybehavior types that are common in the
winning group.
Persistence and Demise of Forager Technology and
Institutions
In our model and simulations, a situation in which most individ-
uals engage in huntinggathering and are either sharers or civics
represents the late Pleistocene forager technological and in-
stitutional order. We now use the model to ask how this forager
distribution of types might endure over long periods, as it appears
to have done for at least 100,000 y before the Holocene, and how
it might be replaced by transitions to a farmingprivate property
social order (in which most individuals are bourgeois farmers)
under the inuence of more farming-favorable Holocene weather
conditions. To do this, we need to determine how the population
share of each of the six technologicalbehavioral types will change
from period to period. We can do this because, for any compo-
sition of a group (conditional on having lost or not lost an in-
tergroup contest), the current fraction of each of the technology
behavior types uniquely determines the expected payoffs of each
type and, hence, given the updating process just described, the
expected change in population frequencies of the types in the next
period. We are especially interested, of course, in those dis-
tributions of each of the types for which the model predicts no
change and that could therefore represent a long period of in-
stitutional and technological stasis.
One of these so-called stationary states is a population com-
posed entirely of civic huntergatherers. To see why this state is
stationary, imagine that by migration or behavioral experimen-
tation a fewbourgeois types should appear in the group. Given the
difculty of demarcating and defending wild resources, bourgeois
types in this setting simply act as aggrandizers, beneting from
others who share the goods in their possession while refusing to
share, and occasionally engaging in contests with civics and other
bourgeois types as a result. [This is consistent with the ethno-
graphic literature and suggested by archaeological evidence (35,
40).] The all-civic state is stationary because the many civics rarely
lose their contests with the few bourgeois types, instead inicting
costs on them. A huntergatherer society with a mixture of civics
and sharers will also be stationary, as long as there are sufciently
many civics to win their contests with the occasional bourgeois
invaders.
If there are few civics, however, aggrandizing individuals will
proliferate even in an economy based on wild resources, so that
a mixed population with both sharers and bourgeois but no civics
is also a stationary state. However, because individual possession
is readily contestable given the huntergatherer reliance on wild
species, huntergatherer groups with a signicant number of
bourgeois individuals are conict-prone, and this reduces their
average payoffs and weakens them in contests with groups com-
posed mostly of civics and sharers. As a result, the foraging tech-
nology with its contestability of individual possession provided an
unfavorable environment for the proliferation of the bourgeois
strategy. This explains why before the improvement in property
rights made possible by farming, sharercivic populations would
have prevailed over populations with a substantial bourgeois
fraction. A consequence is that there would be few bourgeois
individuals in the meta population, resulting in relatively low
levels of within-group conict among foragers. This prediction
from our theory is conrmed in the simulated evolution of our
populations, as we will see.
The introduction of farming alters the population dynamics by
facilitating the delimitation and defense of rights of possession.
The all-civic and mixed sharer and civic population distributions
remain stationary states (as long as there are sufciently many
civics). However, as possession is less contestable for farmed
goods, bourgeois individuals do not challenge each others prop-
erty. As a result, the all-bourgeois state is stationary: In an all-
bourgeois population, ones payoffs are maximized by remaining
a bourgeois. When universally adopted, private property in farmed
goods thus represents an evolutionarily stable institution satisfying
the two desiderata of property rights: providing incentives and
avoiding conicts.
As a result, farming-cum-private property groups could have
had average payoffs at least as great as forager groups despite the
inferior productivity of farming, because these groups endured
fewer costly conicts over possession. This would be more likely to
occur, of course, during the Holocene, at times and places where
for environmental reasons the productivity shortfall of farming
relative to foraging was most attenuated.
This reasoning leads us to expect that, under Holocene climatic
conditions, if a group somehow attained a substantial fraction of
bourgeois farmers, the state would persist over long periods. This
expectation, too, will be borne out in our simulations. However,
the above reasoning does not explain how the critical mass re-
quired for the adoption of both farming and farming-friendly
property rights could have occurred in the rst place. Our next task
then is to use our simulations to show how this may have occurred
and to discover whether the causal model we have calibrated does
indeed track what is known about the Holocene revolution.
Simulated Holocene Transitions
Our simulations reproduce the timing of the best-studied cases of
the independent emergence of agriculture in the archaeological
record (Fig. 2). Equally important, the chicken-and-egg problem
of farming and private property in our model explains why the
transition was a very unlikely event, occurring in only 31 of 1,000
metapopulations that we simulated. The results are robust to
plausible variations in the parameters (SI Appendix).
Note in Fig. 2 that the brief climate ameliorations around
37,00035,000 and 15,00013,000 y ago are associated in the
simulations with short-lived experiments with farming, the latter
coinciding exactly with the well-documented Natuan proto-
farming episode (41, 42). Fig. 3 shows that in the simulations, as in
the archaeological record, mixed farming and huntinggathering
is the norm over very long periods, and that the process of tran-
sition when it occurred was prolonged, highly varied, and some-
times halting.
Process of Transition in the Archaeological Record
Exploring this process empirically is difcult because whereas
farming leaves traces such as specialized tools and genetic and
morphological markers of domestication in plants and animals,
there are no comparable markers of property rights, for which
evidence is as a result necessarily indirect. About storage, for ex-
ample, Kuijt and Finlayson very plausibly write that a transition
4 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1212149110 Bowles and Choi
from extramural to intramural storage system may reect
evolving systems of ownership and property . . . with later food
storage systems becoming part of household or individual based
systems (2). However, one can hardly expect to nd direct ev-
idence for the coevolution of property along with technology
during this period.
Southwestern Asia provides the best-documented cases pro-
viding evidence of the gradual adoption of food production along
with evidence suggesting the emergence of private property in
stores in the Levant between 14,500 and 8,700 B.P. (SI Appen-
dix). At the beginning of this period, Natuans hunted and col-
lected wild species and possibly practiced limited wild-species
cultivation along with limited storage (41, 42). Somewhat before
11,000 B.P. there is direct evidence of storage of limited amounts
of wild plants outside of dwellings, consistent with the hypothesis
that access to stored goods was not limited to the members of
a residential unit (2, 43). A millennium later, goats and sheep
had been domesticated (constituting a substantial investment),
and we nd large-scale dedicated storage located inside dwell-
ings, suggesting more restricted access (4, 44).
Over none of this period could one describe these communities
as either simply foragers or farmers. Their livelihoods were mixed;
in many cases, their residential patterns varied over time between
sedentary and mobile, and it seems their property rights, too,
varied among the types of objects concerned, with elements of both
private and common property in evidence. Bogaard (4) and her
coauthors found that at Catalhoyuk in central Anatolia (10,500
10,100 B.P.), families stored their own produce of grain, fruit,
nuts and condiments in special bins deep inside the house. This
restricted-access storage coexisted with the prominent display of
the horns and heads of hunted wild cattle. The authors concluded
that plant storage and animal sharing was a common juxtapo-
sition for the negotiation of domestic [the authors elsewhere
call it private] storage and interhouse sharing. The process of
change was neither simple, nor monotonic, nor rapid. However, in
both its institutions and its technology, Levantine people were
living in a very different world in 8,700 B.P. from the world of the
early Natuans almost six millennia earlier.
Discussion
In many histories of technology, the key event is the invention; the
subsequent spread occurs inexorably as the result of its superiority
in lessening the toil required to sustain life. This model has been
suggested for the Holocene revolution (45); but it does not work. No
invention was necessary. Kent Flannery, who pioneered archaeo-
logical studies of the emergence of farming, observed that we know
of no human group on earth so primitive that they are ignorant of
the connection between plants and the seeds fromwhich they grow
(46). Moreover, foraging and farming populations interacted over
long periods in the Levant, India, Scandinavia, and elsewhere. In
these cases, those whoremained foragers surely knewabout the new
technology, as did foragers long before the initial spread of farming.
In our simulations, as in the archaeological record, groups with
substantial fractions of farmers coexist over long periods with
groups engaged almost exclusively in foraging.
Nonetheless, the independent adoption of farming was a rare
event. Our coevolutionary model and simulations suggest that
after the amelioration of climatic conditions at the end of the
Pleistocene attenuated the productivity disadvantage of farming
compared with foraging, in a few populations the new technology
and behaviors proliferated synergistically. We thus provide an-
other piece of evidence in support of the views of those archae-
ologists and anthropologists who have advanced the idea that
cultural aspects of the Holocene revolution are essential to un-
derstanding the advance of farming as a new technology (13, 47,
48). The institutional changes required for the exploitation of
cultivars, observed Andrew Sherratt, were a privatization of
resources [that] marked the end of the forager sharing ethic (49).
The key event here is thus not the invention of farming but the
coincidence of sufciently many individuals adopting both the
novel property rights and the new technology so as to overcome
the critical mass problem.
Once established, communities of farmers would eventually
outreproduce foragers due to the lower costs of child rearing as-
sociated with sedentary living (17, 50) even if the productivity of
farming fell short of foraging. The resulting demographic ad-
vantage may help to explain the second part of the puzzlethe
spread of farming and private property once it was reasonably
advanced within a group. However, this Neolithic demographic
transition was protracted, especially in the Levant and other
places where farming was independently introduced, extending
over more than two millennia (51). As a result, the rst groups
where most individuals took up the farmingprivate prop-
erty package would not have enjoyed signicant demographic
Eastern USA
Sub Saharan Africa
Northern South America
New Guinea Highlands
Yangzi and Yellow River Basins
Fertile Crescent
Natufian Proto-farming
40000 30000 20000 10000 0
v
a
r
i
a
t
i
o
n

o
f

1
8
O
%

o
f

m
e
t
a
-
p
o
p
u
l
a
t
i
o
n
s

5
4
3
2
1
5
4
3
2
1
0
0
Measured
climate volatility (left)
Simulated incidence of
private property rights
and farming (right)
years before present
Central Mexico
Fig. 2. Climate variability and the emergence of farming and private prop-
erty. The gray bars (Right) indicate for the year given the number of simulated
populations (of the 1,000 simulated) in which more than 50% of the entire
population were bourgeois farmers. Estimated dates of some well-studied
cases of the initial emergence of cultivation are on the horizontal axis (8, 54,
55). Climate variability (Left) is an indicator of the 100-y maximum difference
in surface temperature measured by levels of
18
O from Greenland ice cores
(SI Appendix). A value of 4 on the vertical axis indicates a difference in av-
erage temperature over a 100-y period equal to about 5 C.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
6000 5600 5200 4800 4400 4000 3600 3200 2800
Fracon of Bourgeois
Fracon of Farmers
Fracon of Bourgeois-Farmers
Years Before Present
Fracon
The Holocene Transion
in a simulated meta populaon
Fig. 3. Simulated process of transition. Shown are the dynamics occurring in
a single implementation of the simulation (a single population representing
an ethnolinguistic unit), indicating (vertical axis) the fraction of the pop-
ulation that are, respectively, farmers, adopters of the new property rights,
and both (bourgeois farmers) for the given year.
Bowles and Choi PNAS Early Edition | 5 of 6
E
C
O
N
O
M
I
C
S
C
I
E
N
C
E
S
advantages for many generations (SI Appendix). Similar compet-
itive advantagesinstitutional, demographic, and (following the
further development of farming methods) economicwere
enjoyed by entire ethnolinguistic units in which most groups had
made the Holocene transition, as evidenced by the encroachment
of Bantu farmers and herders on the territories of foragers in
Africa, and similar expansions of populations of farmers of Eu-
ropean descent virtually the world over.
Our model does not take account of this encroachment on for-
agers by farming ethnolinguistic populations with their formidable
military capacities and political reach, in most cases occurring long
after the advent of the Holocene. Had this not occurred, much of the
world today might look much like the Australia, Western Cape of
southern Africa, or California encountered by the rst Europeans
no more than half a millennium ago: populations of both sedentary
and mobile huntergatherers practicing at most intensive foraging
with limited rights of property based on individual possession.
Thus, it is possible that until improvements in the productivity
of seeds and the food value of cultivars made farming signicantly
more productive, most of the independent Holocene transitions
were not driven by the prior appearance of a superior technology
but instead by a coevolutionary process. We therefore think that
a conventional account of historical dynamicsthat an advance
in technology occurs and institutions followfails as a description
of many of the archaeologically documented early transitions to
farming and its associated property rights. The technology-driven
account fails for two reasons: The new technology was not initially
an advance, and it most likely did not precede the emergence of
the institutions favoring farming as a livelihood.
Our model of the coevolution of institutions and technologies may
nd applications to other epochs that saw the joint emergence of
apparently synergistic institutions and technologies (52). An exam-
ple is the European industrial revolution of the 18th and early 19th
century, which sawthe introduction of steampower and mechanized
production and the reorganization of production around a new
economic institution: the factory and employment for wages rather
than family-based production by independent producers (53). A
further application may be to the challenges to intellectual property
rights posed today by the new information-processing technologies.
ACKNOWLEDGMENTS. We thank Amy Bogaard, Stephen Shennan, Ian Kuijt,
Bruce P. Winterhalder, Michael Gurven, Tom Dillehay, Gregory Dow, Paul
Seabright, Robert Rowthorn, and Henry Wright for contributions to this re-
search; and the Behavioral Sciences Program of the Santa Fe Institute, The
Russell Sage Foundation, and the US National Science Foundation for sup-
port of this project.
1. Kuijt I (2008) Demography and storage systems during the southern Levantine Neo-
lithic demographic transition. The Neolithic Demographic Transition and Its Con-
sequences, eds Bocquet-Appel J-P, Bar-Yosef O (Springer, Dordrecht), pp 287313.
2. Kuijt I, Finlayson B (2009) Evidence for food storage and predomestication granaries
11,000 years ago in the Jordan Valley. Proc Natl Acad Sci USA 106(27):1096610970.
3. Garnkel Y, Ben-Shlomo D, Kuperman T (2009) Large-scale storage of grain surplus in
the sixth millennium BC: The silos of Tel Tsaf. Antiquity 83(320):309325.
4. Bogaard A, et al. (2009) Private pantries and celebrated surplus: Storing and sharing
food at Neolithic Catalhoyuk, Central Anatolia. Antiquity 83(321):649668.
5. Earle T (2000) Archaeology, property and prehistory. Annu Rev Anthropol 29:3960.
6. Bogaard A (2004) Neolithic Farming in Central Europe (Routledge, London).
7. Barker G (2006) The Agricultural Revolution in Prehistory: Why Did Foragers Become
Farmers? (Oxford Univ Press, Oxford).
8. Bellwood P (2005) First Farmers: The Origins of Agricultural Societies (Blackwell,
Oxford).
9. Oliver SC (1962) Ecology and Cultural Continuity as Contributing Factors in the Social
Organization of the Plains Indians (Univ of California Press, Berkeley).
10. Wiessner P, Tumu A (1998) Historical Vines: Enga Networks of Exchange, Ritual, and
Warfare in Papua New Guinea (Smithsonian Inst Press, Washington, DC).
11. Zvelebil M, Rowley-Conwy P (1986) Foragers and farmers in Atlantic Europe. Hunters
in Transition, ed Zvelebil M (Cambridge Univ Press, Cambridge, UK), pp 6793.
12. Gregg SA (1988) Foragers and Farmers: Population Interaction and Agricultural Ex-
pansion in Prehistoric Europe (Univ of Chicago Press, Chicago).
13. Moore AMT, Hillman GC, Legge AJ (2000) Village on the Euphrates: From Foraging to
Farming at Abu Hureyra (Oxford Univ Press, New York).
14. Bowles S (2011) Cultivation of cereals by the rst farmers was not more productive
than foraging. Proc Natl Acad Sci USA 108(12):47604765.
15. Cohen MN, Crane-Kramer G (2007) Ancient Health: Skeletal Indicators of Agricultural
and Economic Intensication (Univ Press of Florida, Gainesville).
16. Cohen MN (1977) The Food Crisis in Prehistory (Yale Univ Press, New Haven, CT).
17. Bocquet-Appel J-P (2009) The demographic impact of the agricultural system in hu-
man history. Curr Anthropol 50(5):657660.
18. Endicott K (1988) Property, power and conict among the Batek of Malaysia. Hunters
and Gatherers, eds Ingold T, Riches D, Woodburn J (St. Martins, New York), pp 110127.
19. Wiessner P (1982) Risk, reciprocity and social inuences on !Kung San economics.
Politics and History in Band Societies, eds Leacock E, Lee R (Cambridge Univ Press,
Cambridge, UK), pp 6184.
20. Woodburn J (1982) Egalitarian societies. Man (Lond) 17(3):431451.
21. Kelly RL (1995) The Foraging Spectrum: Diversity in Hunter-Gatherer Lifeways
(Smithsonian Inst Press, Washington, DC).
22. Dyson-Hudson R, Smith EA (1978) Human territoriality: An ecological reassessment.
Am Anthropol 80(1):2141.
23. Zeder MA (2011) The origins of agriculture in the Near East. Curr Anthropol 52(S4):
S221S235.
24. Zeder MA (2008) Domestication and early agriculture in the Mediterranean Basin:
Origins, diffusion, and impact. Proc Natl Acad Sci USA 105(33):1159711604.
25. Price TD, ed (2000) Europes First Farmers (Cambridge Univ Press, New York).
26. Smith B (1995) The Emergence of Agriculture (Sci Am Libr, New York).
27. Zvelebil M (1986) Hunters in Transition: Mesolithic Societies of Temperate Eurasia and
Their Transition to Farming (Cambridge Univ Press, Cambridge, UK).
28. Bettinger, et al. (2007) The transition to agriculture in northwestern China. Dev Quat
Sci 9:83101.
29. Denham T, Haberle S, Lentfer C (2004) New evidence and revised interpretations of
early agriculture in Highland New Guinea. Antiquity 78(302):839857.
30. Smith B (2001) Low-level food production. Journal of Archaeological Research 9:1.
31. Smith JM, Price GR (1973) The logic of animal conict. Nature 246:1518.
32. Kaplan H, Hill K (1985) Food sharing among Ache foragers: Tests of explanatory hy-
potheses. Curr Anthropol 26(2):223246.
33. Kaplan H, Gurven M (2005) The natural history of human food sharing and cooperation:
A review and a new multi-individual approach to the negotiation of norms. Moral
Sentiments and Material Interests: The Foundations of Cooperation in Economic Life,
eds Gintis H, Bowles S, Boyd R, Fehr E (MIT Press, Cambridge, MA), pp 75113.
34. Boehm C (2000) Hierarchy in the Forest (Harvard Univ Press, Cambridge, MA).
35. Hayden B (2001) Richman, poorman, beggarman, chief: The dynamics of social in-
equality. Archaeology at the Millennium, eds Feinman G, Price T (Kluwer/Plenum,
New York), pp 231272.
36. Wiessner P (2005) Norm enforcement among the Ju/hoansi Bushmen: A case of
strong reciprocity? Hum Nat 16(2):115145.
37. Boehm C (1993) Egalitarian behavior and reverse dominance hierarchy. Curr Anthropol
34(3):227254.
38. Boyd R, Gintis H, Bowles S (2010) Coordinated punishment of defectors sustains co-
operation and can proliferate when rare. Science 328(5978):617620.
39. Boyd R, Richerson PJ (1985) Culture and the Evolutionary Process (Univ of Chicago
Press, Chicago).
40. Pettitt PB, Bader NO (2000) Direct AMS radiocarbon dates for the Sungir mid Upper
Palaeolithic burials. Antiquity 74(284):269270.
41. Unger-Hamilton R (1989) The Epi-Palaeolithic southern Levant and the origins of
cultivation. Curr Anthropol 30(1):88103.
42. Bar-Yosef O (1998) The Natuan culture in the Levant, threshold to the origins of
agriculture. Evol Anthropol 5(5):159177.
43. Willcox G, Stordeur D (2012) Large-scale cereal processing before domestication
during the tenth millennium cal BC in northern Syria. Antiquity 86(331):99114.
44. Kuijt I (2011) Home is where we keep our food: The origins of agriculture and late
Pre-Pottery Neolithic food storage. Paleorient 37(1):137152.
45. Childe VG (1942) What Happened in History? (Penguin, London).
46. Flannery KV (1968) Archaeological systems theory and early Mesoamerica. Anthro-
pological Archaeology in the Americas, ed Meggers BJ (Anthropol Soc Washington,
Washington, DC), pp 6787.
47. Smith B (2011) A cultural niche construction theory of initial domestication. Biol
Theory 6(3):260271.
48. Rosen A (2011) Change and stability in an uncertain environment: Foraging strategies
in the Levant from the Early Natuan through the beginning of the Pre-Pottery
Neolithic B. Sustainable Lifeways, ed Miller NF (Univ of Pennsylvania Mus Archaeol
Anthropol, Philadelphia), pp 128149.
49. Sherratt A (1997) Climatic cycles and behavioural revolutions: The emergence of
modern humans and the beginning of farming. Antiquity 71(272):271287.
50. Lambert PM (2009) Health versus tness: Competing themes in the origins and spread
of agriculture? Curr Anthropol 50(5):603608.
51. Guerrero E, Naji S, Bocquet-Appel J-P (2008) The signal of the Neolithic demographic
transition in the Levant. The Neolithic Demographic Transition and Its Consequences,
eds Bocquet-Appel J-P, Bar-Yosef O (Springer, Dordrecht), pp 5780.
52. Pagano U, Rowthorn R (1994) Ownership, technology and institutional stability.
Struct Change Econ Dyn 5(2):221243.
53. Marglin S (1974) What do bosses do? Rev Radic Polit Econ 6(2):60112.
54. Smith BD (1997) The initial domestication of Cucurbita pepo in the Americas 10,000
years ago. Science 276(5314):932934.
55. Dillehay TD, et al. (2012) Chronology, mound-building and environment at Huaca
Prieta, coastal Peru, from 13 700 to 4000 years ago. Antiquity 86(331):4870.
6 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1212149110 Bowles and Choi
1

Supplementary Information for
The coevolution of farming and private property during the early
Holocene

Samuel Bowles and Jung-Kyoo Choi


Contents

1. Background Evidence
1.1. Productivity of labor in farming and foraging
1.2. Temperature data
1.3. Private property in forager, sedentary hunter-gatherer and farming societies
2. The model of within-group interaction
2.1. Choice of technology: Farming vs. foraging
2.2. Pairing and play
2.3. The sharer and bourgeois strategies
2.4. The civic strategy
2.5. Payoffs
3. Strategy updating with conformism and group contests
4. Replicator dynamics (in a foraging economy)
5. Benchmark parameter values
6. Aggrandizers: A gangster path to the Holocene revolution?
7. Algorithm structure
8. Calibrating farming productivity from climate data
9. Sensitivity of simulation results to alternative productivity calibrations
10. Robustness checks for the remaining parameters
11. Contested private property within groups
12. The Holocene transition in the Levant
13. Demographic advantages of farming and the Neolithic Demographic Transition (NDT)

2

List of Figures

Fig. S1. Late Pleistocene and Holocene climate
Fig. S2. The probability of successfully punishing an individual claiming the entire product
Fig. S3. Biased assignment of models to members
Fig. S4. Within-group dynamics in a foraging economy
Fig. S5. Distribution, group competition, and updating
Fig. S6. Calibration structure
Fig. S7. Simulated distribution of the relative productivity of farming and foraging under early
Holocene conditions
Fig. S8. Climate-induced variation in agricultural productivity relative to foraging
Fig. S9. Simulation runs under alternative assumptions concerning the relationship of climate to
the relative productivity of farming and foraging
Fig. S10. Simulation runs under alternative parameter values
Fig. S11. Fraction of all interactions that are result in contests

List of Tables
Table S1. Mean caloric returns per hour of total labor (c*) for wild and cultivated species with
adjustments for risk and (for cultivation) land abundance and delayed returns
Table S2. Actions taken by the row player when paired with the column strategy
Table S3. Row players payoff
Table S4. Parameter values
Table S5. The advent of farming and the changing nature of storage: Levant 14,500 BP to 8,700
BP
3

1. Evidence.
1.1 Productivity of labor in farming and foraging
Measures of energetic output per hour of labor can be computed for a few forager
populations and for some cultivars farmed under conditions and using technologies that may
have characterized farming at the time of the agricultural revolution (1). Comparison of caloric
returns for foraging wild species and for cultivation requires that account be taken of the
following differences. While seasonal variation in resource availability and delayed return
production are present and storage is possible to some extent by those foraging wild species, the
extent of storage is considerably greater for those who cultivate rather than forage. The reduced
diversity of sources of nutrition among farmers gives rise to far greater seasonal and annual risk.
Estimates must therefore take account of losses during storage, time delay, and risk exposure.
Table S1 below give estimates of the calories per hour of labor (including that involved in
processing and storage). Data are for cereals only due to the lack of the comparable data on non-
cereal cultivars of the first farmers (such as avocado, bottle gourd and squash.)

Table S1. Mean caloric returns per hour of total labor (c*) for wild and cultivated species with
adjustments for risk and (for cultivation) land abundance and delayed returns. (Standard deviations
in parentheses). The estimates relevant to an individual's initial decision to engage in some farming (line 2)
entail no greater risk for the farmer than for the forager. The estimates relevant to average reproductive
output for a group of farmers (line 3) account for the greater risk exposure of farmers. The p-value for the
difference between the wild and cultivated c* distributions are from the Wilcoxon rank-sum test (not
affected by the possibly exaggerated returns in the Great Basin prehistoric data). (The Welch-
Satterthwaite difference in means t-test (unequal sample variances) gives (for the three rows in order): t =
2.33, 2.38 and 2.39. Source: (1)

Estimate
Cultivated
(1)
Wild
(2)
p value
(for wild-cultivated)
Ratio
(1)/(2)
1. No risk or time delay adjustment
1041
(152)
1662
(590)
0.005 0.63
2. Decision: forager risk only and
subjective delay
954
(147)
1628
(578)
0.0003 0.59
3. Evolution: risk and reproductive
delay
951
(139)
1628
(578)
0.0003 0.58

1.2 Temperature and climate data
Temperature data. Differences in temperature (Centigrade) are about 1.2 times the
difference in the
18
O signal shown in Fig. S1 (2). The data indicate that changes in mean
temperature as great as 8 degrees (C) occurred over time spans as short as two centuries. By way
4

of comparison, the Little Ice Age that devastated parts of early modern Europe experienced a fall
in average temperatures of one or two degrees, and the dramatic warming of the last century
raised average temperatures by one degree, comparing the unprecedentedly hot 1990s with a
century earlier (3, 4). The variability of climate during the late Pleistocene required high levels
of geographical mobility, which was an impediment to any substantial investments in tree crops
or field preparation or even stores and storage facilities. The scale and pace of climate change is
truly extra ordinary: for example
18
O signals from sea cores indicate that between 25 and 60 ka,
variations in sea surface temperature of 3
o
5
o
C occurred over periods of 70 years or less in the
Santa Barbra Basin, California (5) (sea surface temperatures today are about this different
between the Santa Barbara Basin and northern Vancouver Island). Think about the frequency of
moves and the distances that early humans may have traveled. A change of 9 degrees Centigrade
in the course of a millennium appears to have been common prior to the Holocene. That's the
difference in the average daily temperature in Cape Town and Mombasa 4 thousand kilometers
to the north. While humans and the wild species on which they depended could of course adapt
to a few degrees change in temperature, we infer that the distances covered and frequency of
moves must discouraged the kinds of investment that farming requires.

Fig. S1. Late Pleistocene and Holocene climate. Greenland ice core data used in the main text

1. 3. Private property in forager and farming societies.
In (6) we provide further evidence for the empirical claims made in the paper and in
particular the claim that the private property rights that farming requiredthe assurance that an
individual or family could exclude others from appropriating the results of their labor in
cultivation and animal tending were for the most part absent among foragers (mobile hunter
gatherers) at the end of the Pleistocene. The sources upon which we have relied are (7-38).
5

The main archaeological and ethnographic evidence is the fact that by comparison with
small scale farmers who can allocate surpluses to privately owned storage, foragers redistribute a
substantial fraction of the food and other economic resources acquired by their own labor to
others beyond the nuclear family, along with the widely held inference that, according to Boehm:
the great majority of prehistoric foragers would still have todays main pattern, meaning that
they lived in mobile, flexible, and egalitarian multifamily units. (7) (p.88)
We have collected evidence on the fraction of the food acquired by family members
(measured in calories) that a family retains for its own use rather than being consumed by others
in both foraging and horticultural societies. The mean of the four estimates of family retention
rates for foragers is 0.295 and of the five estimates for horticulturalists is 0.644. (The difference
in means is significant at p< 0.001.)
But a family can expect that those receiving transfers will reciprocate, that is transfer
something in return as a result of the initial transfer that is above and beyond the amount that
would have resulted from family ties, genetic relatedness, propinquity and other influences on
sharing . A more adequate measure of the amount retained would be 1 (fraction transferred )(1-
reciprocation rate), where the reciprocation rate is defined as the fraction of the quantity
transferred from family A to family B that is reciprocated in transfers from B to A, controlling
for other influences on transfers of food such as geographical proximity and kinship. The means
after adjusting for reciprocation are 0.360 and 0.740 for the foragers and the horticulturalists,
respectively. Sources: Ache (forest) (29); Yora (30); Aka (24, 31); Hiwi (32); Rakoiwa,
Krishisiwa and Bisaasi (1986 and 1987) (33); Yekwana (34).

2. The model of within-group interaction
2. 1. Choice of technology: Farming vs. foraging
An individual may be either a hunter-gatherer or a farmer. A hunter-gatherer obtains I
h
by
foraging. A farmers product is acquired by a prior investment of an amount z, resulting in a
subsequent gross product of (using the subscript a to refer to farming or agriculture) I
u
u
= rz ,
where r is a measure of the returns to the farmers investment that varies inversely with climatic
volatility (see Fig. S8.) The net payoff of the farmer (after subtracting the cost of investment)
would be I
u
= rz z if the product is not taken by others or z if the product is taken by others.
Thus the expected payoff to farming may be less or greater than I
h

depending on climatic
conditions and the prevalent property rights. Note even if the productivity of the two economies
6

is identical (I
h
= I
u
) the farmer is more vulnerable than the forager to losses through hostile
challenges over their product because the farmers crop (rz ) exceeds the foragers prey ( V
h
) by
the magnitude of the farmer's prior investment, z. We set I
h
=1, r =1.5, z=2 (i.e., I
u
u
=3 and
I
u
=1) for our benchmark simulations reported in Fig. 2 in the text. The benchmark returns to
farming investment r =1.5 is adjusted downward according to climatic conditions so that
realized returns are r 0 (see Section 8).

2. 2. Pairing and play
Once a player obtains I
h
or I
u
u

depending on his technology, he plays the bourgeois-sharer-
civic game with a randomly chosen partner twice; one game over the distribution of his product
and the other game over the distribution of his partners product. Each player has one of the
behavioral strategies described below and summarized in Table S2.

2. 3. The sharer and bourgeois strategies
The first behavioral type, the sharer, concedes half of the product to the other, or the whole
product if the other claims it (similar to the dove in the hawk dove game). Bourgeois individuals
claim the entire product if it is in their possession; if not they act like a sharer. A non-possessing
bourgeois may engage in contests with another bourgeois if the possession of the product is
contestable, which occurs with probabilities that depend on whether the product is farmed (p
u
)
or foraged (p
h
). The loser of the contest bears a cost, while the winner claims the product (I).
We let p
h
> p
u
because farmed plants and animals and the land that supports them are more
easily demarcated and defended than most foraged products, due to the greater productivity of
land devoted to domesticated species. In our benchmark model of the forager economy we let
p
h
=1 so a bourgeois will contest any claim of individual possession (under these conditions,
bourgeois is thus identical to hawk in the hawk- dove game).

2. 4. The civic strategy
A civic individual shares when he meets either sharers or other civics. Being paired with
anyone claiming the whole product, all the civics attempt to punish him. They succeed with
probability , in which case the civic obtains the entire product and distributes it equally to all
civics; if unsuccessful, which occurs with probability (1- ), all the civics bear the cost
C
(1-u-[)n
,
7

where n is the number of civics, and o and [ are the within group frequencies of sharers and
bourgeois, respectively (i.e., 1 o [ is the frequency of civics). The probability of civics
successfully punishing an individual claiming the entire product () is increasing in their
numbers (Fig. S2) according to:
f =
|(1 o [)n + u.Sn :]
y
|(1 o [)n + u.Sn :]
y
+ |n (1 o [)n u.Sn + :]
y

where : and y are a center value and exponent of the winning function. The following figure
depicts this probability (when : =8; y =5). In this case, the civic successfully punishes with
probability 0.5 if the number of civic in a group is equal to 8.


Fig. S2. The probability of successfully punishing an individual claiming the entire product.
Successful punishment of a group member claiming the entire product depends on both the numbers
willing to contest the claim and the legitimacy of their objection in the eyes of others. We take the number
of civics in the population to be a measure both of the degree to which the target is outnumbered, and the
legitimacy of the punishment.

Table S2. Actions taken by the row player when paired with the column strategy
Bourgeois Sharer Civic
Bourgeois
If possessor, claims the entire
product, if not concedes it
If possessor, claims the
product, if not concedes it
If possessor, claims the
product, if not concedes it
Sharer
If paired with a possessor,
concedes the product
Shares the product equally Shares the product equally
Civic
If paired with a possessor (who
claims the entire product) seeks
(jointly with other civics, if any)
to contest the bourgeois claim
Shares the product equally Shares the product equally

2.5 Payoffs
With these assumptions, we have the payoff matrices described below. The following payoff
matrices are for the games when a farmer meets a farmer, when a farmer meets a forager, when a
forager meets a farmer, and when forager meets a forager, respectively.
8

Table S3. Row Players payoff
(Recall that when a pair meets, they play two games, one over the product of each of the two
parties. Expressions in red are the rows payoff of the game over rows product and those in
black are the rows payoff of the game over columns product. We assume p
u
=0, p
h
=1, and
I
u
u
= rz )

(a) When a farmer meets a farmer
Bourgeois Sharer Civic
Bourgeois
|I
u
u
z] + |u] |I
u
u
z] + _
1
2
I
o
0
_ |C + (1 )I
u
u
z] + _
1
2
I
u
u
_
Sharer
_
1
2
I
u
u
z_ + |u] _
1
2
I
u
u
z_ + _
1
2
I
u
u
_ _
1
2
I
u
u
z_ +_
1
2
I
u
u
_
Civic
_
1
2
I
u
u
z_ + _
I
u
u
(1 )C
(1 o [)n
_ _
1
2
I
u
u
z_ + _
1
2
I
u
u
_ _
1
2
I
u
u
z_ +_
1
2
I
u
u
_

(b) When a farmer meets a forager

Bourgeois Sharer Civic
Bourgeois
|I
u
u
z] + _
1
2
(I
h
C)_ |I
u
u
z] + |I

] |C + (1 )I
u
u
z] + |C + (1 )I
h
]
Sharer
_
1
2
I
u
u
z_ + |u] _
1
2
I
u
u
z_ + _
1
2
I
h
_ _
1
2
I
u
u
z_ +_
1
2
I
h
_
Civic
_
1
2
I
u
u
z_ + _
I
h
(1 )C
(1 o [)n
_ _
1
2
I
u
u
z_ + _
1
2
I
h
_ _
1
2
I
u
u
z_ +_
1
2
I
h
_

(c) When a forager meets a farmer
Bourgeois Sharer Civic
Bourgeois
_
1
2
(I
h
C)_ + |u] |I
h
] + _
1
2
I
u
u
_ |C + (1 )I
h
] + _
1
2
I
u
u
_
Sharer
|u] + |u] _
1
2
I
h
_ +_
1
2
I
u
u
_ _
1
2
I
h
_ + _
1
2
I
u
u
_
Civic
_
I
h
(1 )C
(1 o [)n
_ +_
I
u
u
(1 )C
(1 o [)n
_ _
1
2
I
h
_ +_
1
2
I
u
u
_ _
1
2
I
h
_ + _
1
2
I
u
u
_

(d) When a forager meets a forager
Bourgeois Sharer Civic
Bourgeois
_
1
2
(I
h
C)_ +_
1
2
(I
h
C)_ |I
h
] + |I
h
] |C +(1 )I
h
] + |C + (1 )I
h
]
Sharer
|u] +|u] _
1
2
I
h
_ + _
1
2
I
h
_ _
1
2
I
h
_ + _
1
2
I
h
_
Civic
_
I
h
(1 )C
(1 o [)n
_ +_
I
h
(1 )C
(1 o [)n
_ _
1
2
I
h
_ + _
1
2
I
h
_ _
1
2
I
h
_ + _
1
2
I
h
_


9

The followings are the explanation on some entries of the payoff matrices.
- The case where an individual (row player) who is a bourgeois farmer meets a bourgeois
farmer (the upper-left cell in Table S3 (a)). The row player keeps his/her product because
the rows farmed product is not contestable and the column player (bourgeois) respects
the rows ownership. Therefore rows net payoff is I
u
u
z. (the first bracket in red).
Because the column player is also a farmer, his/her product is not contestable. The row
player respects his/her partners ownership and receives 0 from the game over his
partners product (the second bracket in black).
- The case where an individual (row player) who is a civic farmer meets a bourgeois farmer
(the lower-left cell in Table S3 (a)). The rows farmed product is not contestable. The
column player (bourgeois) respects the rows possession of the product because he/she
knows that the row is the owner. In response the row player (civic) is willing to concede a
half of the product and enjoy the other half for him/herself. Therefore the row player
receives I
u
u
2 z (the first bracket in red). And the column player refuses to share
because he/she is a bourgeois and attempts to defend his/her whole product. There are
n(1 o [) civics, they successfully punish the column player with probability (see
Fig. S2), in which case all the civics share the column players gross product (I
u
u
since
the column player is a farmer), or lose the fight with probability 1 , in which case they
bear the cost (the second bracket in black).
- The case where a bourgeois farmer meets a bourgeois forager (the upper-left cell in Table
S3 (b)). The farmer (row player) can successfully keep his/her whole product because the
farmed product is not contestable and his/her bourgeois partner would not claim it, which
gives the row player I
u
u
z (the first bracket in red). However the row player claims the
partners product because it is foraged product (note that a bourgeois simply behaves like
an aggrandizer if the product is contestable). Therefore there will be a conflict and both
will get (I
h
C)2 (the second bracket in black).
- The case where a bourgeois forager meets a bourgeois forager (the upper-left cell in
Table S3 (d)). Both will behave like aggrandizers because both products are contestable.
Therefore each one will receive (I
h
C)2.
- The case where a civic forager meets a bourgeois forager (the lower-left cell in Table S3
(d)). The bourgeois forager will attempt to grab the civics product because the civics
product is contestable. All the civics in the group collectively punish the bourgeois for
10

his/her aggrandizing behavior. The collective punishment is successful with probability ,
in which case all the civics share the product among themselves; otherwise they bear the
cost of conflict (the first bracket in red). Since the bourgeois foragers product is also
contestable and the bourgeois refuses to share it, another collective punishment is
imposed on the bourgeois. The collective punishment of the civics is successful with
probability and the civic-forager receives the payoff shown in the second bracket in
black.

3. Strategy updating with conformism and group contests
After all games have been played, each member is paired with a cultural model, possibly a
teacher, religious leader, successful hunter or farmer, or competitor. This pairing process reflects
the way the group socializes its members. If the model and the member are of the same type, the
member simply retains his trait. If the two have different traits, then the member compares his
payoff from this period to the model's payoff, and switches to the model's trait if the model's
payoff is higher.
Members in a group which lost an intergroup contest are paired with a cultural model from
the winning group and update their strategy through the process described above. Between-group
interaction occurs once every generation. An intergroup contest occurs when one of the two is
sufficiently likely to win, reflecting the fact that as with other primates, evenly matched human
groups seek to avoid costly conflicts (39). We assume that when group i is randomly paired with
group j and has a contest with probability d where d=|n

n
]
|(n

+ n
]
). When a contest occurs,
the group with a higher average payoff wins with probability 0.5 + 0.5d. Further we assume that
each member in the losing group loses 3 payoff units, and this score is added to each member in
the winning group.
The pairing rule will introduce conformism to the transmission process if each member of
the more numerous type within a group is more likely to be drawn to be a cultural mode than
would occur by chance. To allow for this, we let the probability that a sharer will be drawn a
cultural model be
o
q
o
q
+ [
q
+ (1 o [)
q

where p (>1) is a measure of conformist biased cultural transmission. The probability that a
bourgeois or a civic is drawn for the cultural model pool is calculated in similar fashion, that is
11

by the same expression but with [
q
in the numerator rather than a
h
and similarly for the civics.
Fig. S3 illustrates the biased assignment of models to members if there are just two types in the
population; for p > 1, the bias is conformist, with larger groups contributing proportionally more
to the pool of cultural models. For p = 1 the pairing of members and cultural models is random.
(For p <1 the bias is anti-conformist, larger groups contributing proportionally fewer to the pool;
we do not consider this case).


Fig. S3. Biased assignment of models to members when there are two types in the population. In the
cultural updating process individuals are assigned to cultural models, but types that are numerically
prevalent in a group are more than proportionally likely to be cultural models (teachers, influential elders,
religious leaders).

4. Replicator dynamics (in a foraging economy)
To calculate the dynamics (without conformism), we assume a payoff monotonic updating
process (behavioral-technology types with higher than average payoffs increase their share of the
population.). We suppose every member in a group is a hunter-gatherer, so that we have I =
I
h
and p = p
h
= 1. Let n
B
, n
S
and n
C
be the expected payoff to bourgeois, sharer and civics,
respectively, and n
S
be the average payoff of the group. Let o, [, and 1 o [ be the within-
group frequency of sharer, bourgeois and civics, respectively. Then we have the following
replicator dynamics.
JoJt = o(n
S
n) and J[Jt = [(n
B
n)

12

where
n
S
= o(I2) + [ u + (1 o [)(I2)
n
B
= [(I C)2 + oI + (1 o [)|(o + [)I + (1 o [)(C)]
and
n = [n
B
+on
S
+ (1 o [)n
C


Fig. S4. Within-group dynamics in a foraging economy. The coordinates of a point in the
simplex sum to 1 and give the fractions of the population composed of the three behavioral types.
For example at point b slightly more than half of the population are sharers and the rest are civics
(there are no bourgeois). The vectors (arrows) indicate the direction of movement in the regions
defined by the loci along which o and [ (the fractions, respectively of sharers and bourgeois),
and 1 o [ (the fraction of civics) are stationary, assuming that the group has not lost a
contest with another group and that conformist assignment of cultural models is absent. The
vectors show that the sharer-bourgeois equilibrium (point d) is self-correcting (asymptotically
stable) while the all-civic equilibrium (point a) is not (it is only neutrally stable, i.e. subject to
drift). Thus beginning at a, drift may propel a population along the left edge of the simplex
towards point b, where the number of civics is just sufficient to repel any invading bourgeois.
Beyond b a group member switching by chance to the bourgeois strategy (or a bourgeois
immigrant) will do better than others in the group, and the population will then be carried to
point d.

The stationary and stable states of this dynamic are the all-civic group and a mixed group of
bourgeois and sharers. These two states give two alternative idealized representation of the
forager livelihood and institutions with the first representing a relatively conflict free social
system and the second representing a breakdown of order with conflict arising over the
possession of valued goods. To understand the dynamics of these societies we need to explore
13

the entire state space. Due to migration among groups and behavioral experimentation there will
always be a few bourgeois individuals present in the group with all civics. Also there will be an
occasional civic in the in a group composed mostly of sharers and bourgeois types. The
population state of bourgeois and sharers is stable (self-correcting) because when civics are few
they often bear the costs of the many contests with the aggrandizing bourgeois in which they
engage (and which they lose because they are few in number). Hence they receive low payoffs
and are not emulated in the updating process. Similarly, when civics are common, the occasional
bourgeois types will not proliferate, as they will lose the many contests with civics that occur,
while civics rarely bear the costs of contests with a bourgeois type (there are few bourgeois so
there is little occasion for punishing, and the numerous civics rarely lose).
But a mixed population of sharers and civics may be unstable because when there are few
civics present, aggrandizing individuals will proliferate, carrying the population to the sharer-
bourgeois stationary state. In our simulations under Pleistocene conditions (that is where
virtually all are foragers) groups composed of a mix of shares and civics occasionally are
invaded by bourgeois types carrying the group to the neighborhood of the sharer-bourgeois state.
But the lower average payoffs in this state due to contestability of foraged products then result in
these groups losing conflicts with groups predominantly made up of sharers and civics, resulting
in the restoration of substantial fractions of civics and sharers and few bourgeois types.

5. Benchmark parameter values

Table S4. Parameter values
Parameter Value Parameter Value
Group size per generation (n) 20 Group interaction Every generation
Migration rate (m) 0.2
Contestability of hunter-
gathered product (p
h
)
1
Behavioral experimentation (e) 0.25
Contestability of Farmed
product (p
u
)
0
Level of conformism (p) 2 Farming productivity (r) 1.5
Hunter-gatherer product ( V
h
) 1 Farming investment (z) 2
Cost of losing a conflict (C) 1.5 Farmers net product ( V
a
) 1

Sensitivity to alternative parameter values is estimated in Section 10. The rationale for these
benchmark parameter values follows. Group size and migration rates are consistent both with
ethnographic and genetic evidence as summarized in (40) and (41). The levels of conformism
14

and behavioral experimentation are not estimated, but appear to be plausible. The hunter-gatherer
product is a normalization. The cost of conflict is chosen to ensure that the underlying game
takes the Hawk Dove form. The rationale for the two contestability parameters is provided in the
text (see Fig. S10 panels B and C for sensitivity to alternative values). Net agricultural product is
set equal to the productivity of hunter gatherers; in the simulations this is endogenously adjusted
downward reflecting the effects of weather volatility (see Section 8 below particularly Fig. S8).
Agricultural investment, which implies a capital to net output ratio in farming of 2 is calibrated
from a farming production function reflecting what for early Holocene Europe is called the
Neolithic package. The capital stock is composed of stored cereals and goats, with the latter
contributing one third of the livelihood of the family, using data on seeds and storage from (1)
and effective lifetime for the livestock of 5 years, possibly resulting in an underestimate of the
capital output ratio for the herding component production. The non livestock portion of the
capital may also be underestimated due to the exclusion of tools. A larger capital output ratio in
farming, by raising the farmers vulnerability to contestation, could be offset by more
intermediate values of the contestability parameters (which would reduce the forager-farmer
differences in vulnerability) without affecting the underlying economic mechanisms at work in
the model.

6. Aggrandizers: A gangster path to the Holocene revolution?
Another path has been proposed, one that proceeds through simple taking and defending of
goods by aggrandizers, who (if their defense is effective) would then have the incentives to
engage in a delayed return activity such as agriculture. This so-called gangster variant of the
Holocene revolution seems quite unlikely, however. We captured its dynamics in the model of
the forager economy in which property rights are always contested p
h
= 1 so that the bourgeois
is just a gangster who claims and fights for resources of others. But in our simulations such
gangsters rarely proliferate and when they do their success is short lived. The reason why the
bourgeois strategy eventually succeeded is that it attenuated conflicts among would be
aggrandizers by recognizing the possession rights of others, once the domestication process
reduced the contestability and contestability of possession. An aggrandizer strategy might
succeed had we assumed strongly asymmetric fighting ability and the inability of those with less
ability to form effective coalitions. But we did not explore this possibility

15


7. Algorithm structure
The code was written in Borland C++ Builder. The simulation algorithm is as follows. A
schematic representation of our model appears in Fig. S5.

A. Create 30 groups and 20 agents in each group.
a. Initially, all the agents are civic hunter-gatherers.
B. Within-group Interaction (bourgeois-sharer-civic game with hunted/gathered product or
agricultural product, see Section 2.)
C. Group Interaction
a. Between-group interaction occurs once every generation. Group i is randomly
paired with group j and has a contest with probability d where d=|n

n
]
|(n

+
n
]
). When a contest occurs, the group with a higher average payoff wins with
probability 0.5 + 0.5d.
b. Each member in the losing group loses 3 payoff units, and this score is added to
each member in the winning group.
D. Updating
a. In a winning group: Each member meets a cultural model, where the model is
chosen non-randomly (p >1) from the winning group.
b. In a losing group: Each member meets a cultural model, where the model is chosen
non-randomly from the winning group.
c. Behavioral experimentation occurs for each individual during the strategy updating
with probability e. If experimentation occurs, the behavioral strategy and the
technology choice are drawn randomly from the entire strategy set. If
experimentation does not occur the individual updates according to the best
response dynamic described above.
E. Migration
a. With probability m=0.2 each individual is selected for a migrant pool from which
individuals are allocated to a randomly selected destination group.
F. Repeat B-E for a specified number of generations.
16


Fig. S5. Distribution, group competition, and updating. Two pre-dominantly forager groups
are represented by the rectangles, populated by individuals who differ in technology (the shape
of the object) and their behavioral type (its color). In the top panels (A) individuals are randomly
paired with others in their group in the distribution stage either conceding, sharing, claiming, or
contesting the claims of others. This determines their payoffs. In the next set of panels (B) the
two groups have had a contest, the left group had higher average payoffs and has prevailed, and
the loser groups members update their type. One updating individual is shown paired with two
models (one for updating behavior, one for updating technology) from the winning group (along
with their payoffs from the distribution stage). Payoffs for all of the pairings shown are given
(with explanation) in Table S3. As a result of the payoff differences, the updating person
converts from a bourgeois farmer to sharer hunter gatherer (shown by the circle in the right hand
C panel). Then (left hand C) members of the winning group (on the left) are paired with models
from their own group, and update their types accordingly. Finally (D) random migration into and
from both groups occurs, along with random experimentation of types (the dashed rectangles).
This sketch does not fully represent our simulation algorithm (e.g. the simulations there are 30
groups each with 20 members of a given generation, and all group members update in every
generation.)


17

8. Calibrating farming productivity using climate data.
The calibration is based on two pieces of information: estimates of the productivity of
farming relative to foraging in the early Holocene (1) and uncontroversial inferences from
climate and other data that conditions during the previous 30 thousand years were exceptionally
farming-unfavorable (42). Among the data supporting the second point (in addition to the
extraordinary temperature variations over less than century long time scales described above in
Fig. S1) is evidence that the total organic carbon stored terrestrially (a measure of photosynthesis)
during the period just prior to the Holocene was 41% of the levels during the Holocene (using
estimates based on the highest available resolution (43)). The two big facts that we represent in
the model are thus that farming, were it to have occurred during the Pleistocene, would have
been very unproductive and would have been more productive during the Holocene but still not
as productive as foraging.
In this section we show two things: (a) that the productivity levels of farming and foraging
generated by the climate data were calibrated to fall within plausible ranges in light of the data
and (b) that the main results are unaffected by plausible variations in the relevant productivity
ratio (see Fig. S9 below).
We used the data on climate volatility to predict movements in levels of farming
productivity in the late Pleistocene and Holocene so as to generate long term averages consistent
with the our direct estimates mentioned two paragraphs above. This calibration then
automatically generated climate-driven temporal variation in the relative productivities over the
entire 40 thousand year period. The average productivity of farming relative to foraging in the
Holocene is based on the data in Table S1 (1). To generate plausible values for this ratio we used
the estimates for caloric productivity of foraging and farming (5 and 15 respectively in number)
to generate 75 pairs representing the productivity ratio for some particular location of a sub
population. The resulting distribution is shown in Fig. S7.
The time series of the ratio of farming to foraging productivity that we used for the
simulations reported in Fig. 2 of the text was generated in the following way (summarized in Fig.
S6).
1. Climate volatility. From the Greenland ice core data we calculated the maximum
difference in the
18
O signal during 5 generations (i.e., the maximum
18
O signal minus
the minimum of
18
O signal during 5 generations) and smoothed the series by using 200-
year (10 generations) backward moving average. For any particular year, this number
18

(which we denote w) is a measure of the climate variability experienced by the previous
ten generations. (Because the NGRIP ice core data are binned in 20 year intervals we also
defined a generation as 20 years.)
2. Farming productivity. We use data on the relative productivity of farming and foraging to
calculate 0, the disadvantage of farming due to the temperature volatility calibrating the
generating function so that on average farming is 86 percent as productive as foraging in
the Holocene and 35 percent as productive during the Pleistocene. For this calibration, we
used 0 = (u.4S w)S for our benchmark simulation.
3. The productivity ratio. In the simulation, the agricultural gross output is rz z where r is
a measure of the productivity of the farmers investment and z is a prior investment for
farming. We apply the disadvantage of farming due to the temperature volatility, 0, to r,
so that the gross productivity of farming becomes (r 0). The productivity of farming
(net of investment costs) is thus: (r 0)z z; and because the productivity of foraging
is normalized to 1, this is also equal to productivity ratio of the farming to hunting.

Fig. S6. Calibration structure.




















Data on Holocene Farming and
Foraging Productivity
Estimated ratio of
productivity of
farming to
productivity of
foraging (p)
Climate Data
Series A
A
Maximum difference
during 5 generations
10-generation average
for smoothing
Disadvantage for
farming (0 )
Temperature
variability (w)
Weather-based
estimate of
productivity ratio (p) :
(r 0)z z
19

Fig. S7. Simulated distribution of the relative productivity of farming and foraging under
early Holocene conditions. Source: methods described in text, data from (1). The arrows
indicate the average productivity ratios generated by our weather data (given by (r 0)z z
where 0 is the weather-determined estimated average disadvantage of farming.)



Fig. S8. Climate-induced variation in agricultural productivity relative to foraging. Shown is the
ratio of the products acquired by farming (assuming that the products are not contested) to the products
acquired by foraging (i.e., I
u
I
h
). This ratio is determined entirely by the weather volatility shown in text
Fig. 2. In the figure p
P
and p
H
are, respectively, the ratio of productivity of farming to the productivity of
the foraging during the late Pleistocene and the Holocene.



9. Sensitivity of simulation results to alternative productivity calibrations
In our baseline simulations the average ratio of farming productivity to foraging productivity
in the Holocene is 0.86 and in the Pleistocene 0.35. We also experimented with simulations
(shown below) in which the Holocene productivity ratio (farming to foraging) was 0.92 and 0.66.
These are shown in Fig. S9.
20

Fig. S9 Panel B shows that a calibration that raises the relative productivity of farming has
limited effect (modestly increasing the number of transitions in the Holocene without affecting
the Pleistocene period). Panel C (remarkably) shows that farming and private property emerge
even when farming during the Holocene is just two thirds as productive as foraging. In this set of
a thousand implementations of the simulation (equivalent to a thousand independent populations)
10 populations made an independent to farming and private property, or about the actual number
of such transitions in the archaeological data. Thus what we call the Holocene revolution could
have occurred despite farming productivity being approximately inferior as our estimates
indicate. Finally Panel D shows a modest increase in the farming productivity ratio that has no
effect on the Pleistocene period.

Fig. S9. Simulation runs under alternative assumptions concerning the relationship of climate to the
relative productivity of farming and foraging. Shown in each panel are for each time period the
number of populations (of the thousand populations simulated) in which half of the population or more
are bourgeois farmers. Panel A is our benchmark for ease of reference. In the figure p
P
and p
H
are,
respectively, the ratio of productivity of farming to the productivity of the foraging during the late
Pleistocene and the Holocene. Source: see text.
A
p
P
=0.35
p
H
=0.86

T
h
e

n
u
m
b
e
r

o
f

m
e
t
a

p
o
p
u
l
a
t
i
o
n
s


B
p
P
=0.41
p
H
=0.92

C
p
P
=0.15
p
H
=0.66

D
p
P
=0.49
p
H
=0.86


Years before present
21


10. Robustness checks for the remaining parameters.

The remaining robustness tests appear in Fig. S10.

Fig. S10. Simulation runs under alternative parameter values. Shown in each panel are for each time
period the number of populations in which half or more of the population are bourgeois farmers of a
thousand simulations of a population under the parameters indicted (all other parameters are at their
benchmark values). These simulations may be compared to the results in Fig. 2 of the main text, which
are reproduced in each panel in black for ease of reference.


years before present

A. Rate of random migration per generation. Higher levels of
migration tend to accelerate the emergence of farming because the
effect is to reduce the force of group competition (because
migration makes groups more similar) and this weakens the process
that stabilizes the forager-hunting technical-institutional
equilibrium. When m = 0.18, there are a total of 14 transitions to
farming and private property at the end of the period. Migration is
random (the so called island model) meaning that with probability
m an individual is relocated in a randomly selected subpopulation
(including the current one). Typically observed migration is highly
non-random (most migrants go to a few nearby sub-populations)
and thus is much less effective in equalizing group compositions
(41, 44, 45). Based on evidence from the Aland Islanders,
the !Kung, and circumpolar Eurasian hunter-gatherers (46) we
estimate that our benchmark random migration rate (0.2)
corresponds to an empirical rate of migration of from 0.3 to 0.6.

years before present

B. Contestability of foraged products, bourgeois strategy. We
assume that farmed species are not contested by non-possessing
bourgeois individuals, but that hunted or gathered species are.
Some hunting and gathering groups establish family based property
rights over wild species (which may be acquired in particular
locations). To explore the sensitivity of our results to this forager
property rights scenario we consider reductions in the contestability
of possession of wild species (that is, partial property rights). The
effect is to facilitate the emergence of private property and hence
farming. Note the property rights first scenario that characterizes
many transitions occurs in our benchmark simulations even under
the least favorable possible conditions, namely that foraged
products are entirely contestable.

22


years before present
C. Contestability of farming products, bourgeois strategy. In
our benchmark simulations we assumed that farmed goods were
not contested by bourgeois individuals; foraged goods were
contested, as were farmed goods in civic-bourgeois interactions,
leading to high levels of property rights contestation in the
benchmark (see Section 11). Here we allow bourgeois farmers to
contest each others private property, with the following
frequencies and resulting transitions.
Benchmark
a
=0, resulting in 31 transitions;
a
=0.05, 10
transitions;
a
=0.10, 5 transitions;
a
=0.15, 1 transition.

years before present

D. Cost of losing a conflict within group. The greater the cost of
losing a conflict the stronger will be the selective forces working
against groups with some bourgeois members but that have highly
contestable property rights. This occurs when groups are engaged
primarily in hunting. As a result, the larger is this cost, the more
beneficial is the adoption of farming (which reduces contestability
of goods and hence diminishes the number of conflicts). As a result
higher values of C result in earlier and more frequent independent
emergences of farming and private property. When C = 1.25, by
the end of the period there are 6 transitions to farming and private
property.

years before present
E. Probability of a between group conflict per generation. In the
simulations group conflicts result in resource transfers and the
assignment of winning group cultural models for the cultural
updating process of the next generation of the losers. Both
contribute to loser populations becoming more like winner
populations. Less frequent group conflicts will weaken one of the
mechanisms stabilizing the forager equilibrium, resulting in both
earlier and more extensive emergences of farming. Based on
evidence of the substantial extent of mortality in between forager
group conflict (47) we think that a once per generation conflict
scenario is not an overestimate. In any case the results are very
similar when conflicts are half this frequent and not qualitatively
affected by even a greater reduction in conflict.
23


years before present

F. Resource transfer from losing to winning group. The larger is
this transfer, the more powerful is the advantage of foraging groups
in competition with farming groups (average payoffs in foraging
groups are almost always higher than in farming groups due to the
productivity advantages of foraging wild species and the costs of
within group conflicts when property rights are contestable). But
the results vary only slightly with variations in this parameter,
suggesting that plausible reductions in the stakes of group
competition have little effect.


years before present
G. Expected waiting time for half of a group of hunters to
become farmers by random experimentation. Random
experimentation creates diversity among groups and over a
sufficiently long time period will convert a group, for example, of
all hunters to a group with equal numbers of farmers and hunters.
We use the waiting time for such a transition as our measure of the
extent of experimentation. In our benchmark case in each period a
hunter will try farming with a probability equal to 12.5 % and for
group of 20 individuals per generation the expected waiting time
for the transition mentioned above is 19292 generations (the
expected waiting time is the reciprocal of the probability that in a
given period 10 or more previously forager group members will
switch to farming by random experimentation, calculated from the
binomial distribution.)

11. Contested private property within groups
Within-group contests over economic resources occur frequently both because punishers
always contest bourgeois attempts to exclude others from the use of resources and bourgeois
contest other bourgeois claims on resources when those resources are contestable (which is
always the case with foraged resources). Fig. S11 shows that the degree of contestation is modest
when bourgeois farmers are either very few or very many; intermediate fractions of bourgeois
farmers which occur during transitions from forager to farmer societies generate substantial
levels of property rights conflict within groups, because (as explained in the text) during these
transitions there is no widely held convention for the management of conflict and the nature of
the economic resources including a substantial fraction of wild species makes private
property claims contestable.

24


(a) all interactions


(b) interactions between bourgeois individuals


(c) interactions between bourgeois and civic individuals

Fig. S11. Fraction of all interactions that are result in contests. Contests occur between bourgeois and
punisher individuals, and between bourgeois individuals when property rights are contestable. Each point
gives average across the 30 groups of the contested fraction of interactions for the population average of
the fraction of bourgeois farmers in the population. The data are from benchmark implementations of the
simulation in which cultivated resources and domesticated animals possessed by a bourgeois farmer are
not contested by other bourgeois individuals. Contestation levels are greater but simulation results are
qualitatively similar in simulations in which contestation of farmed resources by bourgeois individuals
occurs (See Fig. S10 panel C).


12. The Holocene transition in the Levant
Table S5 adapted from (48) presents the data referred to in the text. The evidence for
harvesting and processing wild cereals by the early Natufians is very strong and for cultivation
proper not as strong. Bar Yosef for example refers to Natufians as perhaps the earliest farmers
25

(49), while Unger-Hamilton concludes that the evidence favors the notion that cereals were
being cultivated in the Early Natufian Additional evidence is found in (50-53).

Table S5. The advent of farming and the changing nature of storage: Levant , 14,500 BP to
8,700 BP. Source: From (48) and other sources (54-56) Slightly different dating of the Natufian
periods is used in (54)

Periods Economy Storage Interpretation
Early & Late
Natufian ca.
14500-11700 BP
Intensive collection and some
cultivation of wild plant
resources; intensive hunting of
wild species (esp. gazelle,
possibly communally); no
evidence of husbandry
Indirect evidence
for some very
limited storage.
Relatively
unrestricted
access
Pre-Pottery
Neolithic A ca.
11,700-10,500 BP
Collection and some cultivation of
wild plant resources; possible
domestication of some plants;
intensive hunting of wild species;
no evidence of husbandry
Dedicated storage
silos (wild plants)
outside in public
(and possibly
inside) of residential
units
Relatively
unrestricted
access.
Middle Pre-
Pottery Neolithic
B ca. 10,500
9,250 BP
Collecting and cultivating wild
plant resources; in some places
wide range of domesticates;
hunting, domestication of goat-
sheep
Dedicated storage
outside and inside
residential units;
moderate volume
Restricted
access
Late Pre-Pottery
Neolithic B &
PPNC ca. 9,250
8,700 BP
Reliance on a restricted range of
domesticated plants; hunting,
increased reliance on goats, sheep,
pigs cows.
Dedicated separate
rooms; high volume
Restricted
access


13 Demographic advantages of farming and the Neolithic Demographic Transition (NDT)
Based on fraction of immature individuals in cemeteries dating from the early Holocene, the
reproductive advantages of sedentary living may have raised the rate of population increase from
barely above zero to about 1.3% per annum at its peak (57) followed by declines and subsequent
fluctuation, with a slower tempo of change in the epicenters [of the advent of farming] than in
the secondary transition zones.(54). The NDT most likely witnessed an increase in fertility and
that mortality rose almost in concert with fertility and not after a long delay.(58) The peak rate
was not sustained over the long period and it seems likely that farmers populations grew at the
rate typical of pre-industrial populations or somewhere between 0.1 and 0.2 percent annually.
While capable of growing at well over 2 per cent per annum, forager population (averaged over
long periods) during the Pleistocene may have been stationary it appears to have been positive
26

growth rates appear to have been the case during the Holocene. Thus the NDT might have given
the communities that took up farming and its associated property rights at most an advantage of
about one percentage point (at its peak). The increase in the rate of population growth from
forager to farmer levels appears to have been extraordinarily protracted. But suppose that the
change in growth rates from about 0.3 to 1.3 took place exponentially over just a millennium
(54). Then it would take 382 years before the farmer population exceeded 1.2 of the forager
population, had they had similar populations at the outset.
27

Works cited

1. Bowles S (2011) The cultivation of cereals by the first farmers was not more productive
than foraging. Proc. Natl. Acad. Sci. USA 108(12):4760-4765.
2. Johnsen SJ, Dansgaard W, & White JWC (1992) The origin of Arctic precipitation under
present and glacial conditions. Tellus B41: 452-468.
3. Mann M, Bradley M, & Hughes M (1998) Global-scale temperature patterns and climate
forcing over the past six centuries. Nature 391(April):779-787.
4. McManus J, Oppo D, & Cullen J (1999) A 0.5 million-year record of millennial-scale
climate variabilty in the North Atlantic. Science 283:971-975.
5. Hendy I & Kennett J (2000) Dansgaard-Oeschger cycles and the California Current
System. Paleoceanography 15(1):30-42.
6. Bowles S & Choi J-K (2012) The Origin of Private Property (Santa Fe Institute)).
7. Boehm C (2012) Moral origins: Social selection and the evolution of virtue, altruism,
and shame (Basic Books, New York) p 424.
8. Soffer O (1985) Subsistence Practices and Durationof Occupation in the Late Valdai. The
Upper Paleolithic of the Central Russian Plain, (Academic Press, London, UK), pp 249-
353.
9. Stoffer O (1989) Storage, sedentism and the Eurasian Palaeolithic record. Antiquity
63:719-732.
10. Waguespack N (2002) Caribou sharing and storage: refitting the Palangana site.
Anthropological Archaeology 21:396-417.
11. Peterson N & Taylor J (1998) Demographic transition in a hunter-gatherer population:
the Tiwi case, 1929-1996. Australian Aboriginal Studies 1998(1):11-27.
12. Hart C & Pilling A (1960) The Tiwi of North Australia (Holt Reinhart and Winston, New
York).
13. Schulting R (1995) Mortuary Variability and Status Differentiation on the Columbia-
Fraser Platteau` (Archaeology Press, Simon Fraser University, Burnaby).
14. Bekar C & Reed C (2011) Land markets and inequality: Evidence from medieval
England. in Simon Fraser University.
15. Woodburn J (1982) Egalitarian Societies. Man 17:431-451.
16. Knauft BM (1991) Violence and Sociality in Human Evolution. Current Anthropology
32(4):391-428.
17. Binford L (2001) Constructing Frames of Reference: An Analytical Method for
Archeological Theory Using Hunter-Gatherer and Environmental Data Sets (University
of California Press, Berkeley).
18. Kaplan H, Hill K, Lancaster J, & Hurtado AM (2000) A theory of human life history
evolution: diet, intelligence, and longevity. Evolutionary Anthropology 9(4):156-184.
19. Winterhalder B (1990) Open field, common pot: Harvest variability and risk avoidance in
agricultural and foraging societies. Risk and uncertainty in tribal and peasant societies,
ed Cashdan EA (Westview, Boulder), pp 67-89.
20. Kelly RL (1995) The foraging spectrum : diversity in hunter-gatherer lifeways
(Smithsonian Institution Press, Washington) pp xvi, 446.
21. Dyson-Hudson R & Smith EA (1978) Human territoriality: An ecological assessment.
American Anthropologist 80:21-41.
22. Dowling J (1968) Individual ownership and the sharing of game in hunting socieites.
American Anthropologist 70:502-507.
28

23. Wiessner P (1982) Leveling the hunter: constraints on the status quest in foraging
societies. Food and the Status Quest: An Interdisciplinary Perspective, eds Wiessner P &
Schiefenhovel W (Berghahn, Oxford), pp 171-191.
24. Bahuchet S (1990) Food sharing among the Pygmies of Central Africa. African Study
Monographs 11:27-53.
25. Hawkes K, O'Connell JF, & Blurton-Jones NG (2001) Hadza meat sharing. Evolution
and Human Behavior 22:113-142.
26. Cashdan EA (1980) Egalitarianism Among Hunters and Gatherers. American
Anthropologist 82(1):116-120.
27. Wiessner P (1982) Risk, Reciprocity and Social Influences on !Kung San Economics.
Politics and History in Band Societies, eds Leacock E & Lee R (Cambridge University
Press, Cambridge), pp 61-84.
28. Gurven M, Allen-Arave W, Hill K, & Hurtado AM (2001) Reservation food sharing
among the Ache of Paraguay. Human Nature 12:273-297.
29. Gurven M, Hill K, & Kaplan H (2002) From forest to reservation: Transitions in food
sharing among the Ache of Paraguay. Journal of Anthropological Research 58:93-120.
30. Hill K & Kaplan H (1989) Population and dry season subsistence strategies of the
recently contacted Yora of Peru. National Geographic Research Report 5(317-334).
31. Bahuchet S (1991) Les Pygmees d'aujourd'hui en Afrique centrale. Journal des
Africanistes 61:5-35.
32. Gurven M, Hill K, Kaplan H, Hurtado AM, & Lyles R (2000) Food transfers among Hiwi
foragers of Venezuela: Tests of reciprocity. Human Ecology 28(2):171-218.
33. Hames R (2000) Reciprocal altrusim in Yanomamo food exchange. Adaptation and
human behavior: An anthropological perspective, eds Cronk L, Chagnon N, & Irons W
(Aldine Gruyter).
34. Hames R & McCabe C (2007) Meal sharing among the Ye'kwana. Human Nature 18:1-
21.
35. Borgerhoff Mulder M, Bowles S, Hertz T, & others (2009) Intergenerational
Transmission of Wealth and the Dynamics of Inequality in Small-scale Societies.
Science 326:682-688.
36. Kaplan H & Hill K (1985) Food Sharing among Ache Foragers: Tests of Explanatory
Hypotheses. Current Anthropology 26(2):223-246.
37. Pearson MP (2000) The Archaeology of Death and Burial (Texas A&M University Press,
College Station, TX).
38. Hayden B (1997) The Pithouses of Keatley Creek (Harcourt Brace College Publishers,
New York) p 140.
39. Wilson M, Hauser MD, & Wrangham RW (2001) Does participation in intergroup
conflict depend on numerical assessment, range location, or rank for wild chimpanzees.
Animal Behavior 61:1203-1216.
40. Bowles S & Gintis H (2011) A cooperative species: human reciprocity and its evolution
(Princeton University Press, Princeton) p 254.
41. Gamez SM, Wilkins J, & Bowles S (2012) Cosmopolitan ancestors: Genetic evidence
that prehistoric populations were not small and isolated. in Santa Fe Institute.
42. Richerson P, Boyd R, & Bettinger RL (2001) Was Agriculture Impossible During the
Pleistocene but Mandatory During the Holocene? A Climate Change Hypothesis.
American Antiquity 66(3):387-411.
43. Beerling DJ (1999) New estimates of carbon transfer to terrestrial ecosystems between
the last glacial maximum and the Holocene. Terra Nova 11(4):162-167.
29

44. Rychkov YG & Sheremet'eva VA (1980) The genetics of circumpolar populations of
Eurasia related to the problem of human adaptation. The human biology of circumpolar
populations, ed Milan F (Cambridge U Press, New York).
45. Harpending HC & Jenkins T (1974)!Kung population structure. Genetic Distance, eds
Crow JF & Denniston C (Plenum, New York).
46. Rogers A & Jorde LB (1987) The effect of non-random migration on genetic differences
between populations. Annals of Human Genetics 51:169-176.
47. Bowles S (2009) Did Warfare among Ancestral Hunter-Gatherer Groups Affect the
Evolution of Human Social Behaviors. Science 324:1293-1298.
48. Kuijt I (2011) Home is where we keep our food: The origins of agriculture and late pre-
pottery Neolithic food storage. Paleorient 37(1):137-152.
49. Bar-Yosef O (1998) The Natufian Culture in the Levant, Threshold to the Origins of
Agriculture. Evolutionary Anthropology 5(5):159-177.
50. Bellwood P (2005) First Farmers: The Origins of Agricultural Societies (Blackwell,
Oxford) p 384.
51. Barker G (2006) The agricultural revolution in prehistory: why did foragers become
farmers? (Oxford University Press, Oxford) p 616.
52. P.C.Anderson (1991) Harvesting wild grains during the Natufian as seen from
experimental cultivation and harvest of wild einkorn wheat and microwear analysis of
stone tools. The Natufian Culture in the Levant, eds Bar-Yosef O & Valla FR
(International Monographs in Prehistory, Ann Arbor), pp 521-556.
53. R.Unger-Hamilton (1991) Natufian plant husbandry in the southern Levant and
comparison with that of the Neolithic periods: the lithic perspective. The Natufian
Culture in the Levant, eds Bar-Yosef O & Valla FR (International Monographs in
Prehistory, Ann Arbor), pp 483-520.
54. Guerrero E, Naji S, & Bocquet-Appel J-P (2008) The signal of the neolithic demographic
transition in the Levant. The Neolithic Demographic Transition and its Consequences,
eds Bocquet-Appel J-P & Bar-Yosef O (Springer), pp 57-80.
55. Kuijt I & Finlayson B (2009) Evidence for Food Storage and Predomestication Granaries
11,000 Years Ago in the Jordan Valley. PNAS 106(27):10966-10970.
56. Wilcox G & Storedeur D (2012) Large scale cereal processing before domestication
during the tenth millennium cal BC in northern Syria. Antiquity 86:99-114.
57. Bocquet-Appel J-P (2008) The Neolithic Demographic Transition, population pressure,
and cultural change. Comparative Civilizations Review 58:36-49.
58. Bocquet-Appel J-P (2009) The Demographic Impact of the Agricultural System in
Human History. Current Anthropology 50(5):657-660.